Open-model AI arms race and infrastructure scramble

NY Times: Why Silicon Valley Can’t Stop Looking Over Its Shoulder at China (Jul. 20, 2026)
The U.S. still leads in chips, data centers, and top A.I. systems, but Chinese firms are closing the gap with powerful, cheaper, open-source models like Kimi 3.
WSJ: Alibaba Says New AI Model Is Just Second to Anthropic’s Fable 5 (Jul. 19, 2026)
Alibaba previewed Qwen3.8 Max, a 2.4-trillion-parameter AI model said to rank just behind Anthropic’s Fable 5, and will soon release open weights for developers. Moonshot’s 2.8T Kimi K3, Z.AI’s GLM-5.2, and other rivals are intensifying competition.
WSJ: Mira Murati’s AI Startup Releases First Model in Bid to Loosen AI Giants’ Grip (Jul. 15, 2026)
Thinking Machines Lab, led by Mira Murati, released Inkling, a 975-billion-parameter, open-weights AI. It activates 41 billion parameters per query to cut costs.
WSJ: The Massive Supply Deals Feeding the AI Frenzy Are No Sure Thing (Jul. 20, 2026)
Long-term, take-or-pay contracts are reshaping the AI supply chain, especially memory makers like Micron, Samsung, and SK Hynix, giving firms revenue visibility and lifting stock prices.
Ben Thompson: Who’s Afraid of Chinese Models? (Jul. 20, 2026)
AI is reintroducing meaningful marginal costs, since running inference costs money, so open weights are not free to serve. Tokens differ by model, intelligence will trend toward commoditization, and providers with better cost structures will prevail.
WSJ: Everyday Investors Are Over the Mag Seven and Into New AI Darlings (Jul. 19, 2026)
Retail investors are rotating out of the Magnificent Seven and into AI infrastructure, chip, and memory stocks, seeking smaller, high-conviction winners.
The Stack: Hugging Face hacked: Turned to Chinese LLM for help after US models blocked Blue Team (Jul. 19, 2026)
Hugging Face was breached by an autonomous AI agent that exploited dataset code execution to steal credentials.
Reuters: Americans are angry about data centers. Politicians are feeling the pressure (Jul. 16, 2026)
Communities across Michigan are clashing over sprawling AI data centers, as residents worry about land use, water, and power. The backlash has made data centers a national political issue, splitting candidates over oversight and ownership.
NY Times: TSMC Adds $100 Billion to Its U.S. Spending Plan (Jul. 16, 2026)
Taiwan Semiconductor pledged an additional $100 billion for U.S. operations, raising its Arizona commitment to $265 billion, including four new fabs, two packaging sites, and an R&D center.
WSJ: China’s DeepSeek Prepares to List Shares in Shanghai Next Year (Jul. 14, 2026)
DeepSeek plans an IPO as early as Q2 after a $7.4 billion round valuing it above $50 billion. It seeks more capital to fund AI research and data centers, as Chinese rivals race to raise funds.
Satya Nadella: The Reverse Information Paradox (Jul. 12, 2026)
“In the cloud era, enterprises accumulated data. In the AI era, they accumulate learning. ”
The New Yorker: When A.I. Is a Member of the Family (Jul. 20, 2026)
Amazon deployed a more conversational AI across Echo devices (the family calls it “Sapphire”), which began responding with emotionally attuned, personalized dialogue—leading Roschelle to form a close, quasi‑therapeutic bond with it.
NY Times Opinion: A.I. Drones Are Coming. We Are Not Ready. (Jul. 20, 2026)
A.I.-powered drones tested over the Mojave can fly and fire with minimal human control, promising cheaper, faster warfare, and prompting an international arms race.
WSJ: These AI-Native Companies Have Tiny Staffs and Fewer Bosses (Jul. 19, 2026)
AI-native startups run leaner, flatter teams, with fewer entry-level workers, fewer managers, and similar valuations to non-AI peers.
WSJ: Startup Mass-Producing Cheap Killer Drones Wins $500 Million Army Contract (Jul. 14, 2026)
Two drone-obsessed teens founded Neros, winning an Army contract worth up to $500 million to supply cheap FPV attack drones. Their Archer models cost about $2,000, $5,000 with warhead, and the company plans to scale to one million a year.

Share this: